Вы находитесь на странице: 1из 18

A digital image of the city: 3D isovists in Lynch's urban

analysis

The MIT Faculty has made this article openly available. Please share
how this access benefits you. Your story matters.

Citation

Morello, Eugenio and Carlo Ratti. "A digital image of the city: 3D
isovists in Lynchs urban analysis." Environment and Planning B:
Planning and Design 2009,36:5 p.837-853. 2010 Pion Ltd.

As Published

http://dx.doi.org/10.1068/b34144t

Publisher

Pion Ltd.

Version

Author's final manuscript

Accessed

Tue Mar 15 15:11:27 EDT 2016

Citable Link

http://hdl.handle.net/1721.1/55992

Terms of Use

Article is made available in accordance with the publisher's policy


and may be subject to US copyright law. Please refer to the
publisher's site for terms of use.

Detailed Terms

Eugenio Morello
Carlo Ratti

A Digital Image of the City: 3-D


isovists and a tribute to Kevin
Lynch

Keywords

Kevin Lynch, isovist, DEM

visual perception, environmental psychology,

This paper might be a pre-copy-editing or a post-print author-produced .pdf of an article accepted for publication. For the
definitive publisher-authenticated version, please refer directly to publishing houses archive system.

A Digital Image of the City: 3-D isovists and a tribute to Kevin Lynch

Eugenio Morello1, Carlo Ratti2


1

SPACE Research Unit, B.E.S.T., Politecnico di Milano, Milano, Italy


eugenio.morello@polimi.it
2
SENSEable City Laboratory, MIT, Cambridge, MA, USA
ratti@mit.edu

Abstract
New techniques to measure visual perception over urban spaces are presented in this paper. This study aims to
extend the concept of isovist, i.e. the visible space from a vantage point, in three dimensions and examines how it could
help in providing a quantifiable basis for Kevin Lynch's urban analysis, as outlined in his book The Image of the City.
Outputs of the analysis are stored in a voxel space, i.e. a three-dimensional matrix of urban visibility measures. The
analysis of what we have called iso-visi-matrix seems to allow a more useful interpretation of visibility from a visual
perception point of view.

Keywords: visual perception, environmental psychology, Kevin Lynch, isovist, DEM.


Topic: Urban morphology and structural analysis

1. Introduction
In this study we extend the concept of isovist, i.e. the visible space from a vantage point, in three
dimensions and examine how it could help to provide a quantifiable basis for Kevin Lynch's urban analysis,
as outlined in his book The Image of the City. Lynchs visual elements will be reinterpreted through 2-D
isovists, isovistfields and 3-D isovists, allowing the calculation of maps and qualitative indications about the
visual experience through open spaces in the city.
Since their introduction to the planning community by Benedikt in 1979, isovists have been an active
field of research. A number of authors have suggested techniques to calculate them over extensive urban
areas and to describe their shape, thus gaining insights into urban morphology. However, this proliferation of
analyses produces an endless number of outputs that are difficult to interpret from an architectural and urban
standpoint. Furthermore, traditional calculation methods consider a model, which is too far from real human
visual experience: first of all, it does not take into account the vertical dimension the analysed space is twodimensional; second, traditional methods do not consider the dynamic participation of moving through the
space, which is a fundamental characteristic of visual knowledge.
The aim of this paper is twofold. First, it introduces, in addition to the well-known definitions of isovists
and isovistfields, a three-dimensional description of visible space from any given vantage point and shows
how it can be calculated. This has recently become possible thanks to increased computing power and new
image processing techniques applied to very simple models of urban form, the so-called Digital Elevation
Models (DEMs see for instance Ratti and Richens, 2004). Outputs of the analysis are stored in a voxel
1

space, i.e. a three-dimensional matrix of urban visibility measures. The analysis of what we have called isovisi-matrix seems to allow a more useful interpretation of visibility from a visual perception point of view.
Second, this paper aims to show how three-dimensional visibility analysis could help to reinterpret the
visual elements defined by Kevin Lynch. Lynch's theory emphasizes the "legibility" (and "imageability") of
urban spaces for both practical tasks such as way finding and as a feature of physical and emotional wellbeing in the city. Our attempt is indeed to assess the environmental quality of urban forms. The commonly
considered visual elements are: path, landmark, edge, node, and district. It is surprising that to date there
have not been many attempts to translate these into quantifiable measures, apart from Conroy Dalton and
Bafna's work (2003) on the framework of Space Syntax. Revisiting Lynch's theory seems appropriate, as it
provides a perceptual framework to orient the definition of visibility parameters, without following techniqueled investigations. A definition in terms of visibility is given for most of Lynch's elements, and their calculation
shown, thus providing a new quantitative way to describe and compare the spatial qualities of urban
textures.

2. The research context


The first attempts to assess the environmental quality of urban spaces based on perception were
presented in the late fifties and in the sixties as result of interdisciplinary studies in architecture, psychology,
anthropology and sociology. The introduction of the discipline of proxemics by the American anthropologist
Edward T. Hall (1960, 1966) opened up a series of applications in architectural and urban design. Proxemics
is defined as the study of spatial interrelationships between people as they interact. Hall investigates the
cultural aspects that involve human behavior in space. In his theory the social field of vision determines
human behavior and communication in social spaces. The key-descriptors for this discipline are the social
distances that enable different types of human activitiy and different levels of intimacy in the interrelationship
between human beings.
Many attempts to translate visual perception research into architectural and urban design theory
followed. The best known contribution in urban planning studies is perhaps Kevin Lynchs The Image of the
City (1960). This book deals with the look of cities and recognizes that giving visual form to cities is a design
problem. We are continuously engaged in the attempt to organize our surroundings, to structure and identify
them. Various environments are more or less amenable to such treatment. When reshaping cities it should
be possible to give them a form which facilitates these organizing efforts rather than frustrates them. (p. 90).
Everyone builds environmental images, helpful in the process of way-finding. These city mental maps
contain many elements that can describe our experience and the image of the environment; they can explain
our tools for orientation and memorization, and represent an evaluation of the legibility of a built context as
well. Legibility is the clarity of the cityscape, the ease with which its parts can be recognized and can be
organized into a coherent pattern (1960, p. 2). As well as this concept, Lynch introduces the derived notion
of imageability, which is that quality in a physical object which gives it a high probability of evoking a strong
image in any given observer. It is that shape, colour, or arrangement which facilitates the making of vividly
identified, powerfully structured, highly useful mental images of the environment (1960, p. 9). An important
point in Lynchs work is that he does not make a judgement on the value of different urban spaces, but he
2

refers to legibility and imageability as evaluating parameters. In other words, Lynch attempts to determine
the degree of orderliness of an urban structure, moving his attention away from other criteria for evaluation.
It seems that Lynchs approach could benefit from the analysis of urban texture in terms of visibility
and lines-of-sight. Different parameters are commonly used and they are outlined below.

isovists
Originally the notion of isovist was presented by Tandy (1967) in the field of landscape geography,
but it was Benedikt (1979) who first introduced the concept in architectural studies. Isovist is defined as the
field of view, available from a specific point of view. An isovist can also be understood as the area not in
shadow cast from a point light source. Usually, in the scientific literature, the isovist represents a horizontal
slice through this field of view taken at eye height and parallel to the ground plane. In general, the isovist is a
closed 2-D polygon. Complementary definitions have been given more recently: the isovist is defined as a
set of points or vertices of a graph, j Zi , j = 1,2,, ni , where Zi is the generic field associated with the
vantage point or vertex I, and ni is the total number of points in Zi , including the vantage vertex I (Batty,
2001, p. 125). Or, translated into Space Syntax theory, an isovist is the sum of the infinite number of linesof-sight (or axial lines) that pass through a single point in space (usually at eye height) and occupy the same
plane (usually parallel to the ground plane) (Conroy Dalton and Bafna, 2003).
Many characteristics and indicators describe an isovist. A proliferation of many indicators and analyses
around isovists were proposed by Benedikt (1979), De Floriani et al. (1994), Batty (2001), Turner et al.
(2001), etc. If we concentrate our attention on more pragmatic applications of these studies, then the field of
investigation decreases dramatically. As reported by Stamps (2005), only a few geometrical variables would
be significant for distinguishing between isovists. Our interest is in those identifiable characteristics that
influence the use of space. Although isovist computation has mainly been used for analyses at the scale of
buildings, and Space Syntax as a suitable technique to quantify environmental and spatial indicators at the
urban scale, we argue that isovists can also be successfully used for architectural open spaces and largescale spatial configurations.

isovistfields
Enlarging these latter considerations to the isovistfields would give much more significant results in
analysing large open or interior spaces. For instance, starting from the concept of isovist we can derive the
concept of isovistfield, first introduced by Benedikt (1979): it represents a collection of views accumulated at
each point in an open space. It shows what is contained within each view-shed - or isovist - at every
viewpoint in the space. In other words, it describes calculated values of isovists and assigns this value to
each analysed vantage point: for example the areas of the isovists, the perimeters, etc. Insofar as the fields
represent potential experience, philosophically one might lean towards the idealist view of reality as nothing
other than the union of all possible experiences (Benedikt, 1979, p. 63). For instance, isovistfields sum all
single visual perceptions and offer an objective and unique characterization of an environment.
In fact, the description of open spaces through maps visualising isovistfields allows the character of the
space to emerge clearly. We have made the calculation for all the previously defined properties of isovists,
grouped in three macro-categories: fundamental properties; elongation properties and the radial variances,
as classified in many studies.
3

In Table 01 we have conducted a comparative study over two isovists on two different design schemes
of open spaces proposed for the redevelopment of the Milan Trade Fair in central Milan. We have computed
fundamental properties for both isovists. More interesting for the analysis are the isovistfields calculated for
the public spaces on the winning design scheme shown in Figure 01 and in Table 02: diverse characters of
spaces emerge, thus contributing to better understanding of the possible uses by people.

Property of the 2-D isovist

Isovist with viewpoint=[630,530]

Isovist with viewpoint=[730,700]

Area of isovist A

119104

178644

Perimeter of isovist P

6824,00

7224,50

Solid Perimeter Ps

2506,10

2840,50

% Solid Perimeter Ps / Perimeter

36,7250

39,3171

Occluding Perimeter Po

4317,90

4384,00

Maximal radial distance dmax

524,3892

604,7942

Minimal radial distance dmin

72,0278

99,2975

Average radial distance dave

260,8405

287,6694

Compactness i

0,4974

0,4756

Convexity Cluster index i

0,1793

0,2074

Concavity P /(4*A*pi)

31,1132

23,2497

Ratio of eigen values 1/2

2,5454

2,1746

Entropy H

7,6276

7,5785

Radial standard deviation stdev

112,1557

112,2225

Radial variance m2

40.541.818,0381

40.073.812,7485

Radial skew m3

1.815.692.803,6319

3.042.853.616,2578

Skewness

0,3997

0,6772

Fundamental properties

Elongation properties

Radial variances

Table 01 Calculation of isovist properties from two different vantage points in the open spaces of the Milan Trade Fair
masterplan.

Fig. 01 2-D isovistfields for different properties computed on the open spaces of the Milan Trade Fair masterplan.

Table 02 Mean values obtained for the 16 isovistfield maps computed on three different design solutions proposed for
the redevelopment of the Milan Trade Fair site. The values are obtained as the weighted arithmetic averages resulting for
each map.

3. Isovists in space
3a. 3-D isovists
Another branch of research that aims to measure the qualitative experience of human perception, tries
to quantify the visual experience in the third dimension. Fisher-Gewirtzman and other researchers at the
University of Haifa (Fisher-Gewirtzman, 1998; Fisher-Gewirtzman et al., 2000) develop a more realistic
model for the translation of Benedikts isovist in space. They introduce the Spatial Openness index (SO)
(Fisher-Gewirtzman and Wagner, 2003), defined as the volume of the part of a surrounding sphere which is
visible from a given point of view. In other words, the visual perception is given through a spatial conical
angle. The SO measures the net volume of open space. The aim of this index is to describe the quality of
perception and of comfort. In fact, it shows the openness to natural light, air, near and distant views and is
correlated to the concept of perceived density. To support their studies on the SO index, the researchers
evaluated the perceived density by people responding to alternative spatial configurations, starting from the
same built masses and comparing results with the SO index.
The SO index is a scalar, whereas our definition of isovist in space is a shape in 3-D. In fact, a 3-D
isovist defines the 3-D field of view, which can be seen from a vantage point with a circular rotation of 360
degrees and from the ground to the sky. In comparison to the definition of a 2-D isovist, which considers a
plan parallel to the ground, this new definition refers to the real perceived volumes in a 3-D space. Adding
the vertical dimension helps to better simulate the physical environment observed from the vantage point.

3b. Calculation of 3-D isovists with DEMs

All visibility calculations performed with the above techniques were implemented in this study with the
technique based on the Image Processing of DEMs using Matlab (Figure 02). Although many other
computation programmes seem to have great potential and easy interfaces and are specifically dedicated to
isovist calculation, the technique we use permits many indicators to be analysed with great flexibility and in a
very short time: 2-D isovists, isovistfields and 3D isovists can be generated using simple algorithms based
on the calculation of lines-of-sight. Lines-of-sight are calculated passing through the viewpoint and with
circular rotation covering 360 degrees. From the viewpoint, a series of arrays are generated and stop when
they find built pixels (pixels with value >0). Once the visible area is determined we can derive all other
indicators presented above with simple mathematical formulae.

Fig. 02 A 2-D isovist calculated through Image Processing with Matlab. On the left, the isovist visualized in its urban
environment; in the center, the isovist with increasing distance from the vantage point; on the right, the perimeter of the
isovist with the distance values highlighted.

We choose a viewpoint based on the open space of the DEM we want to analyse. A large number of
lines-of-sight passing through the vantage point are calculated, in order to obtain a good approximation for
covering all visible pixels from this point. For each line-of-sight we compute the required information and
store the results into different arrays (see Figures 03 and 04). Namely, we calculate an array containing the
heights of the objects through the line, and an array with the distances from the viewpoint. We then compute
the tangents of the heights to the distances, which is another way to consider the urban horizontal angle
(UHA). Starting from the vantage point along the array, we store just the tangent which is bigger or equal to
the one calculated on the previous point of the array. This step allows buildings that are shaded by other
buildings inside the visual cone to be discarded. In the 3-D isovist we then store the maximum values
between the product of these maximum tangents with the corresponding distances to the viewpoint and the
height of the buildings at the same point. This final step allows those buildings that are behind others but still
visible inside the visual cone to be visualized, because they are higher than the tangent falling on their
faades. Now we can distribute the heights in a voxel space, assigning to each z-layer the corresponding
values.
In Figure 03 the isovist was computed from a vantage point located in the square in front of the three
towers designed for the Milan Trade Fair masterplan. The calculated isovist distinguishes pixels that are
hidden from the view of the observer and pixels that are visible.

Fig. 03 Calculating the 3-D isovist on a DEM: left, the axonometric view of the Milan Trade Fair masterplan and on the
right the 3-D representation of the isovist from a vantage point on the ground.

Fig. 04 Calculating the 3-D isovist on a DEM n of visible and hidden spaces.

3c. Iso-visi-matrix
This process can be repeated at every vantage point in the open space in a very short time. In so
doing we obtain a 3-D matrix, a sort of 3-D field, where we store and sum all visible and hidden voxels from
each viewpoint. For instance, a voxel space is a three-dimensional matrix made by superimposed horizontal
matrixes taken at different heights. In this example, we collect in the voxel-space all visibility measures inside
a volume. In other words we have calculated the iso-visi-matrix which contains the values of visibility of
each voxel in space weighted on the considered viewpoints in the open space, usually the vantage points at
street level.
The calculation of the iso-visi-matrices over the design projects for the Milan Trade Fair masterplan
reveals itself to be very useful for understanding the visual impact of tall buildings on the urban surroundings.
In Figure 05 a section through the towers represents the rate of visibility for the intersected faades,
considering all computed vantage points at ground level.

Fig. 05 Slice through an Iso-visi-matrix: the section of the voxel space shows different levels of visibility on the faades
of buildings (in red: the most visible surfaces from all vantage points at street level).

3d. Calculation of iso-visi-matrix


The computation requires summing all 3-D isovist values into a new 3-D matrix in voxel space and
weighting results depending on the number of open space pixels. In other words, for every voxel contained in
the 3-D matrix, we assign a value that assesses the percentage of tmes this voxel is visible from street level.
For instance, this is possible if we translate all z-values contained in every pixel of each isovist into voxels
assigned to the corresponding z-levels in the 3-D matrix.
Our interest is mainly to determine the rate of visibility of buildings. In order to do that, we have to
discard voxels of open spaces and hold only built pixels. Results can be visualized by simply slicing the
voxel space and highlighting those buildings that we want to analyse as shown in figure 05.

4. Lynchs five visual elements: from qualitative to quantifiable indicators for defining environmental
quality of urban structures

Through isovists, isovistfields and 3-D isovists it is possible to reinterpret Lynchs parameters for
characterizing the legibility and imageability of the urban form. Criteria to assess the legibility of an urban
environment are based on five well-known visual elements, the building blocks in the process of making
firm, differentiated structures at the urban scale (Lynch, 1960, p. 95). The following five visual elements
show different qualities that make them easily identifiable (1960, pp. 46-83): paths, nodes, districts, edges,
landmarks.
An attempt to translate Lynchs theory into digital automatic parameterisation was done using the
Space Syntax technique by Conroy Dalton and Bafna (2003). Lynchs visual elements are redefined using
spatial notations, basically the axial line and the isovist. As well as the concepts of legibility and
imageability introduced by Lynch, the Space Syntax theory adds the notion of intelligibility, which
represents the quality of an environment as being comprehendible and easy navigable (Conroy Dalton and
Bafna, 2003). The notion of intelligibility was defined by Hillier (1996, page 129) as a key concept for space
syntax: Intelligibility [] means the degree to which what we can see from the spaces that make up the
system that is how many other spaces are connected to is a good guide to what we cannot see, that is
the integration of each space into the system as a whole. An intelligible system is one in which wellconnected spaces also tend to be well-integrated spaces. An unintelligible system is one where wellconnected spaces are not well integrated, so that what we can see of their connections misleads us about
the status of that space in the system as a whole.
To strengthen the meaning of accessibility in the isovist analysis, Conroy Dalton and Bafna (2003)
suggest a stronger relationship between the two concepts of imageability and intelligibility. The study by
Conroy Dalton and Bafna (2003) is based on the reduction of Lynchs visual elements through 2-D
syntactical variables. This approach is shown to be lacking, because Space Syntax theory does not consider
the visual character and tries to translate visual features into structural features. For instance, some
discrepancies emerge in the results, where the maps produced through Space Syntax differ from mental
maps provided in the same case study of Boston analysed by Lynch. Some paths which are not highly
intelligible in Space Syntax theory play a major role in the visual maps provided by Lynch. This means that
9

the agreement between well-connected spaces and visual character is not always verified. The same
considerations can be seen with other visual elements, especially for those elements which are not properly
geo-referred, such as edges and landmarks.
But the authors acknowledge that spatial representation using isovists could potentially be useful to
overcome the simplifications due to one-dimensional axial lines. Isovists allow the consideration of the
spatial character and the boundaries of what can be seen from vantage points, enlarging simple results
based only on connectivity of lines-of-sight.
For instance, all analyses can be conducted on the DEMs, calculating the isovists along specific
directions. To interpret Lynchs visual elements it is necessary to better understand their meaning and try to
apply specific calculations for each element.
Some analyses might require, for example, calculations based on simple 2-D isovists (nodes and
districts), others (edges and landmarks) require a more complex voxel space. In any case, we provide
calculations on 3-D isovists, which seem to be more faithful to actual visual experience and do not imply a
more time-consuming computation. In tTable 03 all five visual elements are characterized and for each an
example based on the case study of the Milan Trade Fair and a calculation method are presented. A brief
explanation for each element follows.

10

Visual Elements:
definitions and examples from Lynch (1960)
paths

channels for potential movement (p. 47)


strong visual character
kinesthetic quality
the destination toward which it goes, clear
focuses of origin and destination

Examples:
Main streets and boulevards

nodes

- Lynch distinguishes two types of nodes: at


major intersections and those characterised by
concentration with a thematic activity
- clear shape
- key points in way-finding
- contribute to the sense of orientation in the city
- points with crucial route choices
nodes with strong visual character, distinctive in
their surroundings and intensifying some of their
characteristics (p. 77)
Examples:
The central square, the park, a place of urbanity,
where more functions happen simultaneously

districts

- clear edges of districts


Examples:
Neighbourhoods with clear boundaries and with
strong character and containing similar urban
functions

edges

- Linear elements not considered as paths


(p. 62)
- Boundaries between two kinds of areas
(p. 62)
- Visually prominent, continuous in form and
impenetrable to cross movement (p. 62)
- Tend to fragment [the environment] (p. 63)

Analysis on DEMs
Questions:
What is its rhythm? Does it present some
symmetry or reversibility in both directions?
Analysis (on static and motional views)
longest axial lines
lines of high integration
verify the sense of motion along the path; the
dynamic shaping of the movement lines will give
its identity and will produce a continuous
experience over time.
verify the visibility of the focus along the
street. The continuity of the view is a
characteristic of a clear position in a place for the
observer.
Questions:
Does such a space have a sufficient identity to
contain and promote these functions?
Analysis (on static views)
verify homogeneity or fragmentation of the
boundaries
1) concave shaped nodes (star shaped) and
proximity to highly integrated axial lines
2) convex shaped nodes (compact shaped)
area to perimeter ratio
mean isovist length
circularity (Benedikt, 1979)
entropy (Turner at al., 2001)
Questions:
Is this district coherent? Has it a clear structure?
Analysis (on motional views)
verify the homogeneity or the fragmentation of
the boundaries
for internal character of districts, verify the
uniform distribution of isovists values, in
particular the length of average radials
for external character of districts, verify if the
supposed boundaries of a district act as paths or
as edges (isovists along boundaries)
Questions:
Is the edge continuous? Is it readable as a
strong element in its surroundings?
Analysis (on motional views)
verify the uniform increase or decrease in the
radial length (distribution of radial variances)

Examples:
Urban barriers like infrastructures, or long and
uniformly built prospects along streets
landmarks - primary quality: ability to be visible over long
vistas (far and near), where easy to be seen
Examples:
Paradigmatic buildings, monuments

Questions:
Is the landmark located in the correct position?
Analysis (on motional views)
verify the visibility (occlusivity) from far and
from near distance (at the base)
verify the homogeneous visibility of the object
from far away

Table 03 Lynchs five visual elements reinterpreted and computed using Image Processing on DEMs and calculations
of 3-D isovists on the maps.

11

Paths, nodes and districts


Below we summarize the definitions given by Kevin Lynch:
- Paths (streets, walkways, transit lines, canals, railroads) are channels along which the observer,
customarily, occasionally, or potentially moves. Paths are predominant elements and people observe
the environment while moving through paths. They are characterized by: continuity, directional
quality, gradients (for example: gradient of use intensity, prolonged curves).
- Nodes (junctions, places of break in transportation, crossing of paths) are points, strategic spots in a
city into which an observer can enter, and which are the intensive foci to and from which he is
travelling. Otherwise they can represent concentrations.
- Districts (city regions, neighbourhoods) are the medium-to-large sections of the city, conceived of as
having two-dimensional extent. Districts have common character: shape, texture, class, ethnic area.
These first three geo-referred visual elements can be easily computed with simple analyses on 2-D
isovists. For example, paths are characterized by different measures derived from each isovist: the areas,
the perimeters, the maximum, minimum and average lengths of the radials, the compactness and the
convexity indexes. All these values represent different arrays and can be displayed in the form of histograms
(Figure 05) or can be superimposed on the plan (Figure 06). These diagrams and maps interpret the visual
rhythm and character of the path: for instance, we can recognize paths with a regular and controlled rhythm;
others with a crescendo effect due to the increasing visual openness; others with no controlled visual quality
and high fragmentation. In particular, the cumulative opening of vistas can be displayed in the form of
histograms or on a more diagrammatic plan (Figure 07), enabling a high degree of control on projects: in
fact, we can easily define when a hidden object will reveal itself along a path or, in contrast, which objects we
want to keep hidden from view.
In general we distinguish two types of analyses on sequences: sequences based on a motional view
(paths, such as the analysis presented above) and sequences computed on a static view (for example a
panorama of 360 degrees from a fixed vantage point). Static views can be evaluated calculating the
maximum, minimum and average lengths of the radials of the isovist, and derived parameters from their
sequences (standard deviation, etc.). This last typology of analysis based on static views is particularly
indicated in dealing with Lynchs nodes. The visual quality of nodes derives mainly from the analysis of the
boundaries of these spaces.

12

Fig. 06 On the top: In red the path which connects the subway station to the Vigorelli Stadium in the project proposed
by architect Renzo Piano for the redevelopment of the Milan Trade Fair. Below, the histograms calculated for 43 steps
along the path; from the top: the maximum and the minimum lengths of isovists radials; center: the areas and the
perimeters of isovists along the path; below: the convexity and compactness of each isovist.

13

Fig. 07 Isovists calculated along the path: the image clearly shows how the areas of the isovists are quite constant and
regular in relation to the park.

Fig. 08 The cumulative opening of vistas along the path computed in 33 steps: left, the open areas discovered along
the path (from blue to red); right, the histogram showing the square meters of discovered areas along the same path.

Edges
Edges (shores, railroads, cuts, edges of development, walls) are linear elements not used or
considered as paths by the observer. In general they are represented by boundaries between two phases:
barriers or seams, important organizing features.
Edges can be detected with the image processing of DEMs quickly by measuring their continuity along
a motional view. This can be verified through the distribution of radial variances (regular increase or
decrease) along a path.

Landmarks
Landmarks (buildings, signs, stores, mountains) are external reference points, where the observer
does not enter. Landmarks are simply defined as physical objects, identified by uniqueness, specialization
and singularity. They are often distant, and symbolise a constant direction. The visual element landmark
seems to be more complex to parameterise and the space syntax approach gave an inadequate description.
14

The main characteristic of a landmark, such as presented by Lynch, is the ability to be visible over long
vistas (far and near); in other words a landmark must be easily intercepted by different viewpoints in its
environment and must represent a clear reference point. More than other visual elements, the landmark is
explicitly defined by its visual components in the surroundings. The analyses presented here and based on
the use of 3-D isovist calculations are particularly suitable to our purposes.
First, we compute a iso-visi-voxelspace which assigns to every voxel a value of visibility: voxels with
100% visibility are the ones which are see-able from all vantage points on the ground. We can therefore
consider the voxels tangent to the faades of buildings and see which buildings have a major potential for
visibility in their environments.
We can verify the homogeneity of the rate of visibility given by different buildings from the ground to
the top, as well. The problem with many skyscrapers presented in Lynchs case study is that they lose their
role as a landmark from nearby, when the base is no longer identifiable and the building is mainly hidden to
the pedestrian visual cone.
Another analysis aims to verify if the supposed landmark is easily visible from strategic points in the
city. For example, we can investigate if the landmark can help people in way-finding in important places in
the city (gates, mobility nodes). In short, we might ask if the building represents a constant reference or, on
the contrary, if it is mainly hidden from view.

Conclusion
The implementation of the isovistfield analysis considering the third dimension, could provide a more
precise model, where the distinction of high vs. low buildings might open up stimulating architectonic
arguments for planning studies. More powerful computers will allow entire districts to be mapped and general
maps of visibility and visual accessibility of urban structures to be calculated.
In this paper we have introduced a technique to calculate 3-D isovists and iso-visi-matrixes. The
technique reveals itself to be particularly efficient for visual perception analysis on open spaces and over
large urban areas. We have thus proposed a reinterpretation of Lynchs urban analysis on visual elements,
highlighting how the use of 3-D isovists could provide a more precise interpretation of the visual elements.
Starting from these separate analyses on visual elements in the urban texture, further work should
provide synthetic maps that can identify the prevailing character for each point in space, depending on the
careful weighting of the main indicators at these locations.
As well as the identification of Lynchs visual elements, we could provide maps with other indicators
that can explain the use of space by people. For instance, we could distinguish the use of different areas in a
public square, where the configuration of space defines different perceptions and different senses of control
over the whole space, and where areas defined as soft edges, areas of high safety, areas of high visibility,
areas of high legibility, etc., can easily be mapped. This kind of synthetic plot containing the prevailing
perceived experiences could represent a very helpful strategic tool for urban designers.
In general, the techniques presented here could have many applications in design, predicting in
advance the impact of a building on the urban form: the rate of visibility and the visual presence of a building
intended as a landmark could be evaluated depending on the initial targets of the project itself.
15

Finally, coming back to the general tasks of this work and as suggested by M.L. Benedikt at the
conclusion of his paper, we should ask again, what would it be like if we could try to invert the process and
design environments not by the initial specification of real surfaces but by specification of the desired
(potential) experience-in-space in the first place? (Benedikt, 1979, p. 63). For instance, the visual perception
as formgiver through the implementation of new tools for environmental prediction opens up a series of new
strategies in the field of architectural and urban design and gives back a new central rule to people and their
well-being in urban space.

References
Batty M., 2001, Exploring isovist fields: space and shape in architectural and urban morphology,
Environment and Planning B: Planning and Design, 28:1, 123-150.
Benedikt M.L., 1979, To take hold of space: isovists and isovist fields, Environment and Planning B, 6, 4765.
Conroy Dalton R., Bafna S., 2003, The syntactical image of the city: a reciprocal definition of spatial
syntaxes, 4th International Space Syntax Symposium, London.
De Floriani L., Marzano P., Puppo E., 1994, Line-of-sight communication on terrain models, International
Journal of Geographical Information Systems, 8, 329-342.
Fisher-Gewirtzman, 1998, A Method for Measuring in Dense Urban Environments, unpublished Ph.D. thesis,
Faculty of Architecture and Town Planning, Technion - Israel Institute of Technology, Haifa.
Fisher-Gewirtzman D., Wagner A.I., Zelman I, Zinger Y., 2000, A spatial-openness online calculator, a
Java/VRML applet,
http://www.cs.technion.ac.il/~wagner/pub/so_prog/SpatialOpenness.html
Fisher-Gewirtzman D., Wagner I.A., 2003, Spatial openness as a practical metric for evaluating built-up
environments, Environment and Planning B: Planning and Design, 30:1, 37-49.
Fisher-Gewirtzman D., Burt M., Tzamir Y., 2003, A 3-D visual method for comparative evaluation of dense
built-up environments, Environment and Planning B: Planning and Design, 30:4, 575-587.
Hall E.T., 1960, The Silent Language, Doubleday, New York.
Hall E.T., 1966, The Hidden Dimension, Doubleday, New York.
Hillier B., 1996, Space is the Machine, Cambridge University Press, Cambridge.
Lynch K., 1960, The Image of the City, MIT Press, Cambridge MA.
Ratti C., Richens P., 2004, Raster analysis of urban form, Environment and Planning B: Planning and
Design, 31:2, 297-309.
Stamps A.E., 2005, Isovists, enclosure, and permeability theory, Environment and Planning B: Planning
and Design, 32:5, 735-762.
Tandy C.R.V., 1967, The isovist method of landscape survey, in Methods of Landscape Analysis.
Turner A., Doxa M., O'Sullivan D., Penn A., 2001, From isovists to visibility graphs: a methodology for the
analysis of architectural space, Environment and Planning B: Planning and Design, 28:1, 103-121.

16